Hi Ilja — some observations and questions on your paper. This is very much from the perspective of the Royal Mail type design that will be introduced into the UK. However I have also referred to the second of our published papers called Stability and Fairness (sounds familiar?) which can be accessed here:
We have to remember that the two papers were written against a background of IDC at experienced in the UK at the time — ie prior to the Freedom and Choice agenda, so they assumed an IDC member would purchase a lifetime annuity. This second paper developed the basic CDC scheme design, and concluded:
More specifically, we have refined our rather simplistic investment model from the first paper now to reflect the age characteristics of each individual plan member, with younger members having a higher weighting to equity assets and older members a higher weighting to bond assets. In conjunction with this, we have adjusted the effect of any one-off adjustment to benefits (positive or negative) so that they are based on a scale of adjustments that reduces with age, thereby exposing older members to less risk of changes (up or down) to their benefit level. Together, these amendments significantly add to the stability of the CDC plan design under very varied membership conditions.
This looks remarkably close to your own description of a fair scheme!
Some observations and thoughts on your paper.
Page 2: CDC schemes are a hybrid form of DB and DC schemes.
Not under the UK model. CDC is a form of DC only with no DB elements. We have worked hard to ensure this, since otherwise employers could be deterred by reclassification risk
Page 3 We define a pension scheme to be fair if all participants, at any point in time, make an arbitrage-free return on the market risk they bear. This definition implies that if a scheme is not fair, some participants could get a better risk-return trade-off outside the scheme. And Page 6 Definition 1 A pension scheme is fair if all participants, at each point in time, make an arbitrage-free return.
I can see that this offers the potential to carry out the analysis you deal with very well in the paper. But it is hardly a real world condition since members rarely have the option to arbitrage against their scheme. Employers will rarely pay the equivalent contribution to another plan of the members choosing. Royal Mail will not, as afar as I am aware, pay any contributions in respect of members who opt out of their scheme, and this would be the norm for all employer sponsored scheme in the UK. In the case of a RM type scheme, with implicit age related employer contributions the calculation of the equivalent pension would be non trivial in any event, and would in theory change over time.
The deffiniton of fair, that we tried to measure through our modelling of the RM scheme, was basically two fold:
- the target benefits were consistent with the contributions payable — to ensure that we were not misleading people
- the incidence of “bad events” was suitably small. Inevitably there is an element of subjectivity in deciding what a bad event is. For most people, a common definition of a bad event is a cut in pension. This is the sort of thing that politicians fear. In practice, if pensioners experience a cut of say 3% in their CDC pension it is not the end of the world. Other aspects of their income — notably state pensions — will continue to be indexed
Page 3 Fairness will depend crucially on the combined specification of the benefit adjustment process and the discount rate process that are used. In this paper we focus on market risk only.
If our method of using best estimate returns to derive the discount rate is inconsistent with your market risk criterion, it would be good to think how your analysis can move on to deal with this approach, since it is most likely to be used in practice in the UK. In our earlier paper we descried now we had derived our discount rate.
The discount rate used in the CDC plan assessment of liabilities for each member is taken as:
– the yield on long-dated fixed interest government bonds, plus
– an equity risk premium in respect of that portion of the liabilities expected to be backed by UK equity holdings at each age into the future, consistent with the plan’s dynamic investment strategy (to make some allowance for expected outperformance of equities over government bonds).
In practice the equity risk premium would be re-calibrated to a suitable best estimate each year by the plan’s actuary based on current market conditions. Our modelling uses a simplistic formula to attempt to capture the first order impact of this re-calibration, with a cap of 5% p.a. and a floor of 0% p.a. applied to the resulting equity risk premium before use in the discount rate.
In this example we did not use the full yield curve — we did for the RM modelling. It was really quite challenging to think how to model the equity risk premium. Each of the three firms had a different approach. I will not ascribe them to the firms!
- Constant long term real equity return. But then when you have “regime change” a different constant.
- Dividend discount model using growth rates based on future inflation (which could be pulled from the fixed and index linked bond markets)
- A variable equity risk premium over the bond yield, using aspects of reversion (via a dividend yield) and moves to a long term trend, based on economic growth. Perhaps with caps and collars.
Okay so perhaps it’s clear this latter approach is ours. It would be good to show how the modelling plays out. If you fix a constant ERP over bonds — say between 3 and 7% — then you get some obvious skews between the cohorts. I will see if I can persuade RM to put the modelling into the public domain.
Page 6 As equity risk and interest rate risk are priced in the market, we also refer to these risk factors as market risks.
I’m still not clear if you means the equity risk premium here. I’m sure you said this was not observable.
Page 7 If policy makers, by contrast, prefer to allow the CDC scheme to redistribute risk in a market inconsistent manner, our fairness criterion will still be useful as a benchmark. Our analysis will make the redistribution flows within the scheme transparent.
This is where I would encourage you to think how your analysis could be applied to the RM design, with its use of best estimate returns to set the discount rate. It may be we need to simplify and think about modelling how the equity risk premium is derived at any point in time
Page 8 The regulatory present value of benefits is determined by discounting all future payments assuming that benefit levels remain constant in the future and using a potentially horizon- dependent ‘regulatory discount rate’. Page 9 Figure 2. Periodic resetting of the regulatory funding ratio F = A in a CDC scheme
In the UK CDC schemes the adjustment will take place continuously (or close to continuously) by virtue of the way in which cash equivalent transfer values are derived, so that assets = liabilities at all times
Page 12 We will keep our specification of the financial market general. By doing so, our evaluation of the fairness of the pension scheme does not rely on subjective assumptions about unobservable or potentially unobservable variables, such as risk premiums or asset price volatilities. We assume that an arbitrage-free financial market exists.
Whether or not an arbitrage free market exists, I come back to the point that the UK version of CDC will use subjective assumptions about future returns , or equivalently future equity risk premiums.
Page 19 The CDC scheme chooses an interest rate sensitivity for its assets that is well below the interest rate sensitivity of the present value of the retirement benefits.
Would this be the case for a scheme with 100% exposure to equities?
Page 27. Technically, if the CDC scheme wants to use a discount rate with deterministic spread, it could maintain fairness by adding a correction term to the benefit adjustment process. And in your conclusions. We show that deviating from a default-free market rate requires the scheme to add corrections terms to its benefit adjustment process
It feels to me — although I freely admit to not trying to do the maths — that the UK approach could come close to this.
- We do not use default free rates, but rather best estimates, which can be thought of as risk free plus a premium.
- We use a benefit adjustment process that looks at the average duration of the scheme and then applies a constant per annum adjustment to benefits. This means the adjustment has much less impact for say a 90 year old pensioner than a 25 year old active member — in other words what you are calling a correction term to the benefit adjustment process.
My question to you would therefore be — is this process fair?
Happy to discuss further
Regards
Kevin
Our second paper called stability and fairness
